Compact flexible actuator based on a shape memory alloy for shaping surgical instruments

1.1 Motivation

Non-invasive and minimally invasive surgery (MIS) present themselves as promising alternatives compared to conventional open surgery due to a lower infection risk and shorter time spent post-op in a medical facility [1]. These advantages are achieved by using special instruments during surgery that can be inserted into the body of the patient through either natural orifices (e.g., mouth, anus) or a limited amount of small incisions. Such instruments have the purpose to provide visual imaging to the medical experts and let them dexterously perform surgical actions such as cutting and grasping within the body cavities. Simple flexible actuators and sophisticated continuum robots can both extend the user’s scope and operation space even in difficult-to-access operation sites. That includes the navigation within the human body, e.g., inside the winding organs of the gastrointestinal tract, as well as in neurosurgery, where critical structures must be avoided by using more complex paths around them [2].

Common actuation methods of continuum robots include pneumatic, hydraulic, electromagnetic, and tendon drives [3]. Continuum robots are numerically controllable manipulators with an infinite number of joints [4] and can be customized to adapt to a specific task. Their design and control remains challenging due to their hyper-redundant characteristic [5] where the number of degrees-of-freedoms (DOFs) exceeds the number of individually controllable actuators. Facing this challenge, aforementioned approaches afford bulky setups such as fluid pumps and valves for pneumatic and hydraulic systems or multiple electric motors for tendon pulling. Consequently, a desired scaling of the robotic structure toward smaller diameters, greater lengths, and larger bending angles is limited. Thus, for usage of controllable and flexible instruments in MIS these aspects of design and performance remain key.

Smart materials such as shape memory alloys (SMA) seem promising when facing those challenges due to their ability to memorize an almost arbitrary shape when traversing a reversible crystal phase change from martensite to austenite caused by increasing temperature, referred to as the shape memory effect. If a task requires one particular target pose including a specific shape of the robotic structure, the SMA actuators can be prepared accordingly in advance (shape setting) and reach that shape in a controlled manner during the actual task execution using simple electric power circuits with little space demand. SMA in the form of wires can provide a higher power-to-weight-ratio than conventional electric motors, they are simpler to implement than other smart materials such as electroactive polymers (EAPs) and allow for larger displacements than piezoceramics [6]. SMA wires are deployed within actuators as quasi-elastic components or as dampers [7] due to their superelastic strain. Furthermore, they can provide linear motion, making use of the SMA for repeatable (workable) strains of up to 5 % [8] when heated via conduction or via Joule heating. Moreover, the linear relation between electric resistance and strain enables SMA actuators’ self-sensing-potential for dual-use as an actuator that is able to sense its position and even further ambient parameters [9]. Therefore, SMAs show great potential for simplified, integrated, scalable and customized actuation of various robotic applications.

1.2 Related work

A widely used actuation principle for continuum robots is the tendon drive, where tendons are pulled by at least one rotatory [10] or linear motor [11] per DOF for each segment, eventually leading to bulky and complex hardware setups. To minimize space requirements, tendons could be replaced by SMA wires or spring coils, which create pulling forces by contraction. Despite the aforementioned large strains, SMA wires must come at great lengths to achieve sufficient bending. Besides those physical bending angle constraints, electronic connectors are required on each end of every SMA actuator [12], leading to a complex, error-prone electronic setup with increased initial load, which is additionally hard to integrate into a delicate device. To compensate for the bending constraint, Allen and Swensen [13] propose a hybrid of SMA and pneumatic actuation, reaching bending angles > 180 ° at the cost of an additional setup for air pressure supply. In Ref. [14], SMA wires are deployed as ‘free-sliding’ tendons within a polymeric substrate and allow for bending angles of up to 400°. When combining multiple of such segments in a serial configuration by rigid links, a hyper-redundant robotic structure occurs, where multiple DOFs can be actuated independently allowing for various complex robotic shapes promising high dexterity and precise controlling of every segment pose [5]. When it comes to the interconnection between those segments, SMAs could be deployed in a straight manner [15], within a pulley system, and in a purely antagonistic configuration, with an elastic joint or a hybrid combination [16]. When including the SMA’s self-sensing ability, even closed-loop control regimes can be realized [17], e.g., for 3D-manipulation of a needle tip in real-time. For further sensing modalities, Ren et al. [18] present a flat soft actuator from elastomer for variable stiffness applications, including embedded SMA and curvature sensors.

1.3 Contribution

The immediate utilization of the shape memory effect has the potential to overcome several existing challenges, such as extending the bending range by creating a suitable shape memory independent of excessive longitudinal strains, to reduce space requirements compared to conventional actuation approaches with large external manipulation and simplifying the electric actuation setup drastically. In this work, we aim to investigate this potential by deploying two SMA wire loops with electrical contacts in proximity and with different shape memories alongside each other. Thus, we provide a proof-of-concept of a novel, shape changing actuator that is able to perform controlled bending and straightening motion for dexterous tasks during MIS.

2.1 Design

The novel actuator consists of two SMA wire loops from nickel-titanium-copper (NiTiCu) forming a skeleton that provides individual actuation once heated above their austenite finish temperature A _f = 65 °C. They are embedded in a monolithic flexible backbone and surround a central supply channel that connects proximal and distal ends of the actuator. The channel provides space for instruments such as fiber optics or bendable needles. The actuator’s proximal end allows for insertion of instruments and to attach the contact electrodes for (controlled) power supply. Several spacer discs serve as wire guidance along the actuator to avoid short circuits. No further cable connection is required for power or control actuation. For this work, two different instrument configurations were investigated and fabricated, as illustrated in Figure 1. Configuration (I) for a bending and (II) for an S-shape of the actuator. Both shape changes are independently obtained by an active-antagonist configuration, as described in Ref. [16]. Thus, each configuration features one SMA wire loop memorizing a shape change (bending or S-shape, respectively) referred to as shape loop (active). The other loop sets back the actuator to its initial straight shape, referred to as straight loop (antagonist). This design allows for a permanently retained pose of the distal tip without further heating if no external force is applied.

Figure 1:

Concept of the compact flexible actuator for minimally invasive instruments featuring two shape memory alloy (SMA) wire loops with preset shape memory for shape changing (orange), reset into straight shape (green) and central supply channel, e.g., for a gripping tool. The desired shapes are bending (I) and S-shape (II).

2.2 Fabrication and assembly

Fabrication steps are presented in Figure 2. In step 1, segments of SMA wire with a diameter of 0.7 mm are cut with a length of 230 mm and cold-formed manually into a U-shape (loop) with a diameter of 3 mm. To set the shape memory in step 2 for straight shape, bent shape and S-shape, heat treatment at 425 °C for 30 min and subsequent quenching in water at room temperature is required for the loop segments [19, 20]. Thereby, each loop segment must be clamped firmly in the desired shape and without torsion stress as it tries to restore its previous shape memory, i.e., the straight shape. Thus, holding racks are fabricated from aluminum, providing the desired curvature of the shapes to stabilize the loop segments during heat treatment. After heat treatment, all wire loops are cold-formed into straight shape at room temperature in step 3. In step 4, the flexible monolithic backbone with intermediate spacer disks and central supply channel is 3D-printed using a stereolithography printer with elastic resin (Formlabs GmbH, Germany). To allow for accurate evaluation, the flexible backbone model is extended for a mounting component at the proximal end and plane elements at each spacer disk for marker attachment. The wire loops are inserted into the printed backbone before its final curing. Thus, the fully cured backbone holds the SMA wires firmly throughout actuation.

Figure 2:

Fabrication steps. Cold-forming wire loops (1), followed by heat treatment with customized aluminum racks at 425 °C (2). All wire loops are straightened by cold-forming (3) and inserted in the 3D printed monolithic backbone, followed by the final curing of the actuator (4).

3.1 Actuation setup

The actuator is mounted at its proximal end in a horizontal position and connected to a laboratory power supply. Once electrified, joule heating increases the temperature of the SMA wires, resulting in the wires memorizing their shape leading to an active shape change, and an antagonistic reset, respectively.

3.2 Visual tracking and actuator shape

To control and characterize the flexible actuator, shape estimation of its body is required. For visual tracking, a RealSense™ D435 camera (Intel Corporation, USA) observes the x-y-plane, tracking nine ArUco-markers attached to the actuator body, as presented in Figure 4a and c. A tracking algorithm is implemented in Python and provides a tracking frequency of 30 Hz. Thus, the positions of the geometrical marker center points provide time-depending shape information. Its performance was evaluated beforehand by tracking a stationary set of markers as utilized on the actuator over a time period of 90 s, revealing a deviation of ±0.35 mm and ±0.05 mm in x and y, respectively. The pose for configuration I is represented by the distal end’s bending. For simplicity, the two markers closest to the distal tip are used to define the actuator’s orientation as the angle φ relative to the x-axis.

3.3 Control

To manipulate the flexible actuator, a PID-controller is proposed in a closed-loop regime as shown in Figure 3 with the iteratively identified control parameters K _P = 50, K _I = 0.5 and K _D = 0.8. The control variable is the bending angle φ of the distal tip. Further shape modeling parameters are currently not considered. For implementation, an ATmega328 (Arduino) microcontroller is deployed using two DC/DC-converters TSR-3 Series (Traco Electronic GmbH, Germany) that can perform each 3 A switching regulation via remote on/off input connected to the pulse-width-modulation (PWM) outputs of the microcontroller at 10 Hz. The computed angle φ is transmitted from the Python script to the microcontroller via Serial Port.

Figure 3:

Closed-loop control setup including marker tracking, angle computation using a Python script and a PID-Controller running on a microcontroller.

3.4 Evaluation methods

Each configuration passes n = 10 experiment runs for workspace characterization, with 180 s cooling pauses in between. For configuration I, we additionally conducted a control evaluation and a qualitative force validation. To characterize the full workspace of the actuator, the shape loop is initially heated at current I = 3A (U _max = 2.6 V) from initial straight shape and the displacement of all markers is monitored over 15 s. A (passive) cooling period of 30 s follows, during which the power supply is switched off. Then, the straight loop is heated for another 15 s. After a second cooling period of 30 s, the experiment is stopped. The cooling periods allow for investigation of possible recoils in scenarios without power supply. Before the next experiment run starts, the actuator is manually cold-formed to its initial straight shape. For force measurement, a spring scale (Micro-Line 20001 1 N, Pesola AG, Switzerland) is insulated at its interface and attached to the distal tip of the actuator (configuration I). Then, I = 3A is applied, and the resulting force is measured. The control regime is demonstrated by following predefined setpoints from straight to bending shape, i.e., φ _ref from 0° to 10° in iteration steps of 2°, each lasting for 15 s.